Hi! My name is Robert Bain. I'm a computer vision and software engineer with an M.S. in Artificial Intelligence. My work spans deep learning, robotics, semiconductor microfabrication, and custom electronics.
We developed a pairwise learning framework that improves data efficiency in image regression and classification. Using a lightweight neural network trained from scratch on a single consumer GPU, we achieved state-of-the-art facial age estimation performance on UTKFace. Submitted to Machine Learning with Applications, 2026.
We developed a deep learning approach to predicting cognitive performance from structural brain MRI. Combining pairwise learning with systematic brain-region elimination, we investigated how to achieve accurate predictions, identify interpretable neuroanatomical features, and maintain practical computational requirements. Research preprint, 2026.
Why are they called artificial neural networks? This paper answers that question by summarizing the influence that biological neural networks have had on ANNs as they became our best perceptual systems in silico. It also documents my attempt to continue the tradition of using neuroscience to empower ANNs.
Paper on developing 3D object detection algorithms with Deep Convolutional Neural Networks.
Paper on visualizing the effects that pruning has on deep neural networks' loss landscapes. Desktop demo: here.
Simulations of hippocampal replay in a spiking neural network path planner.
Marjan Rashidi and I trained a VQ-VAE on 2D slices from T1-weighted MRI scans, achieving strong reconstruction performance on unseen brain data. We used the quantized outputs for downstream tasks like binary classification, demonstrating the model’s utility for feature extraction and anomaly detection.
2nd page 3rd page 4th page 5th page 6th page 7th page 8th page
